Dual-Process accounts claim that responses to reasoning tasks often default to automatically cued belief-based responses. However, recent findings show that when participants are instructed to evaluate the believability of a conclusion, its logical status interferes with their judgment. This finding is inconsistent with the view that belief based judgments are cued automatically. In this paper we present the results of three experiments that examined the impact of a secondary task (random number generation) on belief and validity judgments. Experiment 1 examined simple modus ponens arguments, experiment 2 included disjunctive syllogisms and experiment three employed a blocked presentation design. In line with previous research belief judgments took longer and resulted in more errors than validity judgments. However, in general, RNG impacted more on validity than belief based judgements. These finding suggest that both belief and logic judgements require effortful processing but draw upon different types of executive resource.